William Dembski

From SkepticWiki

Contents 1 Definition 2 Complexity and information 3 Is CSI measurable? 4 Dembski's version of the argument from design 5 Is Dembski doing mathematics? 6 Related topics

[edit] Definition

William Dembski is a mathematician and proponent of Intelligent Design

[edit] Complexity and information

Dembski's key argument for Intelligent Design is that living things are too complex, or, to put it mathematically, have too high an information content, to exist without an intelligent designer. In his words:

"I will argue that the defining feature of intelligent causes is their ability to create novel information." (Dembski, No Free Lunch)

His chosen measure of information is called "Complex Specified Information", abbreviated to CSI. We should note that this is not a measure used by mathematicians: it is not equivalent to Shannon's measure of information, or Kolmogorov's[2]. Indeed, Dembski writes:

"It is CSI that within the Chaitin-Kolmogorov-Solomonoff theory of algorithmic information identifies the highly compressible, random strings of digits." (Dembski, No Free Lunch)

So it seems that things have CSI if they have low Kolmogorov complexity, and it follows that there is more CSI in the letter "a" repeated a million times than there is in the complete works of Shakespeare. We may also note that the sequences of "letters" in a sequence of DNA has a very high Kolmogorov complexity, and so, if we are to take Dembski seriously, it has very low complexity by his own measure.

Dembski also claims that CSI is inversely related to probability:

"Complexity and probability therefore vary inversely --- the greater the complexity, the smaller the probability." (Dembski, No Free Lunch)

This raises several problems. One is that a thing may be highly improbable, and yet have high Kolmogorov complexity: the pattern made by a random sequence of coin tosses is one such example. Nor does such a sequence show intelligent design. It is true that if you toss a coin a hundred times, then the probability of you getting that particular sequence is vanishingly small; yet we would not in ordinary language say that this sequence is high in "specified information", and even if we did, we would not come to the conclusion that this specified information requires an intelligent designer, since we know perfectly well that it was produced at random.

There are also things which fit Dembski's requirements of low probability and low Kolmogorov complexity, and which are not designed. For example, consider the Mandelbrot set. It would seem to have high CSI, for its probability, considered as a picture from the set of possible pictures, is low. Moreover, it only takes a very short computer program to specify the picture, so it has low Kolmogorov complexity, which Dembski says means high CSI. And it certainly looks complex. Yet in spite of all these things it has no designer: it is an artifact of mathematics.

To take an example from biology, consider the "fairy rings" formed by some species of fungus. As an arrangement of mushrooms, out of possible arrangements, it has low probability. The closer they are to regular circularity and spacing, the lower the Kolmogorov complexity, and the higher the CSI. The reason they are called fairy rings is that before any causal explanation for this was known, the regularity of the rings was attributed to intelligent (and supernatural) agency. And if Dembski's inductive argument for design holds, they were quite right to do so. For might not the believers in fairies have repeated, after Dembski:

"In every instance where the complexity-specification criterion attributes design and where the underlying causal story is known ... it turns out design actually is present; therefore design actually is present whenever the complexity-specification criterion attributes design."

Now in this case, the fairy rings are complex by the measures that Dembski has proposed, and once they seemed so complex that people ascribed them to fairies rather than to natural laws. They might well have reasoned as Dembski reasons, and yet we know that Dembski's argument for design fails in this case, for the correct explanation for fairy rings involves the workings of nature, and not fairies.

[edit] Is CSI measurable?

The other problem with CSI is how to measure "probability". In the case of tossing coins, we can do so. In the case of living organisms, this is more difficult: for the whole point about evolution is that it provides a mechanism whereby what appears to us to be complex about organisms, namely their coadaptation of parts, and ultimately of genes, is rendered highly probable. This difficulty is shown up by his criticism of Richard Dawkins' simplistic computer model of one aspect of evolution, which uses mutation and selection to turn random strings into the phrase "METHINKS IT IS LIKE A WEASEL":

"As the sole possibility that Dawkins's evolutionary argument can attain, the target sequence in fact has minimal complexity (i.e. the probability is 1 and the complexity of the usual information measure is 0."

It seems to follow that if evolutionary processes work --- if they guarantee complexity in nature (in the sense of the form and function) then there is actually no complexity in nature in the sense of CSI. It follows that in order even to estimate the CSI of an organism, we must first make up our minds about whether it evolved! But this, of course, is the very question that Dembski wishes to solve by claiming that living creatures have high CSI.

We might wonder what Dembski would make of the opinions of his fellow-Christian, the distinguished paleontologist Simon Conway Morris (famous for his work on the Burgess Shales). Morris believes that life evolved by natural selection and that this is part of God's plan: he has written a book on the subject with the subtitle The Inevitability of Man. In that case, human beings would have a CSI of zero, just like Dawkin's phrase about weasels --- and humans would be the product of intelligent design. So in that case, Dembski's "design filter" would rule out an Intelligent Designer even though he existed.

And this leads to a more general problem: Dembski's idea of whether a thing has CSI seems to vary according to how the thing was produced: for example, Dawkins' phrase "METHINKS IT IS VERY LIKE A WEASEL". According to Dembski, this has no CSI when produced by Dawkins' algorithm, because it was the inevitable result of the algorithm. But it is surely true that when Dawkins himself came up with the phrase "METHINKS IT IS VERY LIKE A WEASEL", he could have chosen any phrase he pleased: his choice was one out of the infinite possibilities of the English language; so when Dawkins wrote (or "intelligently designed") the phrase, it had very high CSI, being very improbable. So the very same phrase can have CSI of zero, if it is produced by an evolutionary algorithm, or infinity, if selected by an intelligent person from the infinite range of English sentences. If we need to know by what processes a thing was produced in order to know whether it has CSI, then we cannot use CSI to tell us how a thing was produced, because if we don't know how it was produced, we can't measure its CSI.

The problem of knowing whether snowflakes have CSI is a case in point. Dembski claims that snowflakes do not exhibit any CSI. This seems a strange claim, since the uniqueness of snowflakes is notorious, and therefore any particular snowflake would seem at first to be high in CSI. However, Behe argues that they do not have CSI because "such shapes form as a matter of necessity simply in virtue of the properties of water". But if Dembski doubted the naturalistic explanation that he gives for snowflakes, and supposed that God or Jack Frost must personally design each snowflake, then he would have to say that snowflakes have high CSI. It seems that in order to "detect" CSI, it is first necessary to make your mind up as to whether the object in question has a designer. Evidently Dembski has made up his mind. Thus, when he says that living organisms have high CSI, and snowflakes have low CSI, he is merely saying that he has made up his mind that the complexity of snowflakes is produced by natural causes and the complexity of living organisms is not. His attribution of CSI to some things and not to others is merely a way of stating in his own unique mathematical jargon that which he believes to be true.

In summary, it seems that Dembski's much-vaunted method of detecting design relies, crucially, on having knowledge of whether or not the thing in question was designed, which makes this method useless for the purpose for which it was intended.

[edit] Dembski's version of the argument from design

Dembski writes:

"In every instance where the complexity-specification criterion attributes design and where the underlying causal story is known (i.e. where we are not just dealing with circumstantial evidence, but where, as it were, the video camera is running and any putative designer would be caught red-handed), it turns out design actually is present; therefore design actually is present whenever the complexity-specification criterion attributes design."

By "circumstantial evidence" we presume that he means the sciences of genetics, geology, paeleontology, comparative morphology and biogeography --- that is, the only evidence we could have for events in nature that took place before recorded history. Very well, let him discount it. Let us also discount for him the speciation events which have been observed in the laboratory and in the wild. Let us ignore all the computer programs, codes, mathematical proofs, circuit designs, chip designs, and so forth which were produced by genetic algorithms. Let us even grant him all that.

Let's look instead at the very things to which Dembski wishes --- devoutly --- to apply his "complexity-specification" criterion. According to Dembski, design may be defined as follows:

"(1) A designer conceives a purpose. (2) To accomplish that purpose, the designer forms a plan. (3) To execute the plan, the designer specifies building materials and assembly instructions. (4) Finally, the designer or some surrogate applies the assembly instructions to the building materials." (Dembski, No Free Lunch)

Now consider the production of a living creature: a tiger, let's say. Presumably Dembski would indeed attribute design to a tiger. The "underlying causal story" of how tigers get made is well-known. What with the prurient interests of the makers of nature documentaries, the video cameras very often are running. And when we observe the process "where any putative designer would be caught red-handed", we do not see any of the four things which Dembski says constitute design. We see ... we see two tigers that love one another very much. We see reproduction with variation, no designer, and no miracles. At this point, an evolutionary biologist might well offer an inductive argument of his own, if scientists did that sort of thing.

As we do not like inductive arguments, we shall follow this parallel no further, but confine ourselves to pointing out that when the video cameras are rolling, we see, in fact, no sign of any of the four things that Dembski identifies as design.

As for the need for intelligence in intelligent design, Dembski's account of the role of intelligence is that it works: "by actualizing one among several possibilities, ruling out the rest, and specifying the one that was actualized". Now, this is almost exactly what evolution does: mutation is like the brainstorming of the different possibilities: natural selection corresponds to ruling out the bad ones. Dembski's intelligent designer seems superfluous, since the same task is achieved by natural processes.

Even if we leave evolution out of it, Dembski's claim that "design actually is present whenever the complexity-specification criterion attributes design" is false. We have given the example of fairy rings: they have low probability among all possible arrangements of fungi (hence high CSI); they have low Kolmogorov complexity (hence high CSI); and people did indeed attribute them to design. And they were wrong. In the same way, the patterns made by frost on window panes has low probability (for the set of all possible patterns) and because of its self-similarity has low Kolmogorov complexity; and yet those who attributed these patterns to an intelligent designer called Jack Frost were in error.

Finally, let's refer to snowflakes again. We doubt very much that anyone has tracked the development of a snowflake as it falls through the air; or whether, if anyone has done so, they were in any position to see whether they were being produced by physical laws or the unseen hand of God. More intruigingly still, there are triangular and twelve-sided snowflakes the existence of which, for the moment, physicists can't quite explain. Yet Dembski is confident that snowflakes are produced by natural processes.

[edit] Is Dembski doing mathematics?

It is noteworthy that although Dembski says he is doing "information theory", his chosen "measure" of "information", CSI, appears nowhere in the mathematical literature. This is not surprising when you consider that in order to "measure" the CSI of something, you need to do empirical research into whether it has a designer. This is not a mathematical concept. Dembski does not publish his work in peer-reviewed journals of mathematics, and has claimed that he prefers to disseminate his ideas in non-peer-reviewed media:

"I've just gotten kind of blasé about submitting things to journals where you often wait two years to get things into print. And I find I can actually get the turnaround faster by writing a book and getting the ideas expressed there. My books sell well." (Dembski interviewed in the Chronicle of Higher Education, December 21, 2001).

We have never heard of any other mathematician refusing to publish his work in the peer-reviewed journals merely on the grounds that there was a long time to wait. Other mathematicians are bound by the academic law of "publish or perish" and queue up to get published: this is why mathematicians have to wait so long for publication.

In fact, the reason Dembski can't publish his work as maths is because it isn't maths. As he writes:

"I'm not and never have been in the business of offering a strict mathematical proof for the inability of material mechanisms to generate specified complexity." [3]

This quotation appears in a response to someone who has pointed out that his maths is faulty. [4]. It is good of him to admit it, but we feel that he should stress this point more often when addressing creationists. For example, when Rob Koons described Dembski as "the Isaac Newton of Information Theory"[5], he was probably under the impression the Dembski had proved something --- perhaps even more than one thing --- in the field of information theory. How disappointed he would be to discover that Dembski has proved nothing whatsoever in that particular branch of mathematics, and that Dembski says that it's not his business to try.

It is also is a curious statement for a mathematician to make. He is, or should be, precisely in the business of providing strict mathematical proofs for things. He has dressed up his ideas in the language of mathematics; and his profession is to provide proofs for mathematical conjectures. Surely the next step should be obvious to him; and one wonders why he doesn't make it his business to find a proof for his claims. Unless and until he provides such a proof, it is premature of him to describe the conclusions he's jumped to as being a law of nature (the "Law of Conservation of Information", he calls it, just as if he was doing real science). This is what professional mathematicians laughingly refer to as "proof by vehement assertion"[6].

So far from proving anything, all that Dembski has done is to restate the Argument from Design in the language of mathematics. This does not increase its academic respectability any more than writing the equation I = E(C) supports the theory that I am the Emperor of China. As he lacks both a logical proof for his mathematics and an empirical proof that his assumptions model the real world, he is as far away from proving the Argument from Design as I am from ascending the Chinese throne.

To take one example of how he rewrites his faith in mathematical jargon, we have seen how when he says that snowflakes have "low CSI" he means no more than that in his opinion they have no designer. To take a fresh example, he uses mathematical functions to model natural processes, and deduces his so-called "Law of Conservation of Information": but in order to do so, he needs one crucial assumption:

"This argument [...] treats functions as mere conduits of information, and does not take seriously the possibility that functions might add information."

That is, his "law of conservation of information" is derived from an assumption that no information can be added by natural processes. But this is to assume the thing to be proved.

It is not clear, then, why Dembski wishes to rephrase in his cumbersome and private mathematical language such commonplace creationist arguments, if he does not intend to use this framework to produce proofs of his contentions. If we were to attribute unworthy motives to him, we might suggest that the principal result of this senseless jargon is to give people the impression (a) that mathematics supports creationism (b) that Dembski deserves comparison to Isaac Newton.

[edit] Related topics

• People
• Argument from design